Lightening the Load: How AI is driving the future of exoskeletons

In our most recent episode of Lexicon, we sat down with Dr. Hao Su , an Associate Professor in Mechanical and Aerospace Engineering at North Carolina State University and a Joint UNC/NCSU Biomedical Engineering Department faculty member.

Join us as we discuss Dr. Su’s work on artificial intelligence (AI) and exoskeletons and reveal the technology’s potential for revolutionizing locomotive medical care in the future.

Bringing humans and robots together with exoskeletons

“Our research is focused on human-centered robotics and AI,” Dr. Su explains.

“We have clinical collaborators from leading rehabilitation hospitals and do clinical testing with people with disabilities. Through these experiments, we found that tuning control parameters takes quite a lot of time, around 20 minutes to half an hour,” he added.

This is far too long in Dr. Su’s estimation and one of the motivating factors behind his team’s work to develop a more efficient system. Another major challenge Dr. Su explains is the adaptability of exoskeleton controllers.

“The great thing about the controller is that it can assist with various activities,” says Dr Su. “The amazing thing is that if the person does walking, running, or stair climbing, the controller can be made adaptive to multiple locomotion activities,” he added.

Building on this, Dr. Su and his team have also developed an AI-powered simulation framework called “Learning in Simulation” to overcome these hurdles.

Learning in simulation

The best part of this is that it doesn’t need to conduct any human experiments, which are very time-consuming. Moreover, it can also be used for a whole range of different locomotion activities.

This framework integrates three interconnected neural networks to simulate human movement, muscle coordination, and exoskeleton control.

One of the standout features of Dr. Su’s simulation framework is its ability to learn control policies. “Typically, many simulation systems don’t consider how to learn a control policy. They only simulate a human itself,” Dr. Su notes.

“In our simulation, the virtual human learns to walk, and the robot figures out how to coordinate with the human. These two learning processes take place simultaneously,” he added.

Dr. Su explained that this high-fidelity simulation includes highly detailed models of humans and robots and their physical interactions. This sets it apart from other state-of-the-art simulation systems that often only focus on simulating humans or robots.

Not just theory

But it is not just theory. Dr. Su and his team have used the simulation framework’s real-world applicability, using a hip exoskeleton. Dr. Su elaborates, “We tested human muscle activity using EMG sensors to measure how the robot can reduce muscle activity and save human effort. We found muscle activity reduction in both simulation and real life.”

The outcomes were promising. “We found that the robot can reduce about 24% metabolic cost during walking, which is equivalent to reducing about 11 kilograms of body mass,” Dr. Su explained to IE.

This significant reduction in metabolic cost was also observed in running and stair climbing, with 13% to 14% reductions.

Not one-size-fits-all

But, as Dr. Su explained, it doesn’t offer a one-size-fits-all solution. However, to help handle the differences in height and biology between the sexes (and individuals), Dr. Su’s team employed a cutting-edge method called domain randomization to account for the nuances in human physiology.

“In the simulation, we provide some disturbance to each muscle to incorporate the variability from many people, which is exciting because it means we can test the limits of what the system can do,” Dr. Su said.

“This makes it incredibly versatile and generic,” he explained. This approach ensures the control policy is adaptable to many individuals, making exoskeleton technology more inclusive.

What is in store for exoskeletons in the future?

It doesn’t end there, either. Dr. Su’s vision extends beyond assisting people with disabilities. “We want to try our learning simulation framework with different robots and people,” he says. His team is currently working on a project sponsored by the National Institute of Health to study exoskeletons for children with cerebral palsy.

Looking ahead, Dr. Su wants to open-source his team’s simulation framework and control policies. “We’re excited to share our software and control policy with the community,” he says.

This move will democratize the technology, allowing more researchers and developers to contribute to its evolution and application across various domains.

Dr. Su’s lab is also working to make his “smart” exoskeletons leaner and lighter. “We are developing an incredible new generation of exoskeletons that are about 30% smaller than existing ones,” he reveals.

If achieved, this should make them more comfortable, less intrusive, and more affordable, making them accessible in home and workplace settings.

Exoskeletons: not just medical devices

“We’re thrilled to be working with BMW in South Carolina to explore how exoskeletons can benefit workers, particularly in upper limb assistance,” he shares. The goal is to drastically reduce musculoskeletal injuries and improve efficiency in tasks such as overhead assembly.

He explained that Dr. Su is also working on ensuring that exoskeletons are user-friendly and adaptable to individual needs. “Our learning simulation control policy is an amazing one-size-fits-all solution, and we have components for human feedback reinforcement learning, too,” he said.

This feature allows users to adjust the assistance level to their preference, enhancing the overall user experience.

Looking to the future

Dr. Su is dedicated to integrating AI and exoskeleton technology into daily life. “Our mission is to make assistive robots available for everyone, for everyday activities,” he said.

This vision involves advancing technology to make these devices more affordable and accessible to a wider audience.

Dr. Su and his team plan to continue pushing the boundaries of what’s possible, making the future of exoskeleton technology look brighter than ever. This promising future brings much hope and an improved quality of life.